Skip to content

SKILLS-TREE.md

Latest commit

 

History

History
115 lines (91 loc) · 5.69 KB

SKILLS-TREE.md

File metadata and controls

115 lines (91 loc) · 5.69 KB

Codebase Context Skill — Skills Tree v2.0.0

For agents: Read this file first. Match your task to a skill below, then read only skills/<name>/SKILL.md for full instructions. Do not read all skills — pick only what you need.

Quick Reference

# Skill Category Model Tags Depends On Token Est. Parallel
1 init context opus index, scan, setup, architecture — ~8K no
2 status context haiku check, health, staleness init ~1K yes
3 query context haiku search, preview, lookup init ~1.5K yes
4 refresh context opus rebuild, update, incremental init ~5K no
5 plan workflow opus plan, task, dependency init ~4K yes
6 build workflow sonnet implement, create, feature init, plan ~6K no
7 fix workflow sonnet debug, bugfix, root-cause init ~5K no
8 refactor workflow opus restructure, blast-radius init ~5K no
9 team workflow opus multi-agent, parallel, coordination init ~8K no
10 test quality sonnet test, verify, auto-fix init ~4K yes
11 audit quality opus security, performance, a11y, dead-code init ~6K yes
12 review quality opus code-review, style, logic init ~5K yes
13 research quality opus docs, errors, best-practices — ~3K yes
14 branch git sonnet branch, switch, context-ref init ~2K no
15 pr git opus pull-request, summary, blast-radius init, branch ~4K yes
16 merge git opus merge, conflict, resolution init, branch ~4K no
17 diff git opus diff, impact, dependency-chain init ~3K yes
18 sync git sonnet pull, push, rebase, conflict init ~3K no
19 log git haiku history, commits, cross-ref init ~1.5K yes
20 stash git haiku stash, wip, restore init ~1.5K yes
21 deploy ops opus deploy, checklist, env-vars, breaking-changes init ~4K no
22 track ops haiku session, task-log, changes — ~1K yes
23 connect setup sonnet mcp, server, endpoint — ~2K no

Categories

Context Management

init · status · query · refresh

Core indexing and lookup. Run init first on any new codebase — all other skills depend on the .ccs/ index it generates.

Workflow

plan · build · fix · refactor · team

Task execution with dependency tracking. plan before build. fix for bugs. refactor for structural changes. team for multi-agent parallel dispatch.

Quality

test · audit · review · research

Verification and knowledge. These skills are read-heavy and parallel-safe — run multiple simultaneously.

Git

branch · pr · merge · diff · sync · log · stash

Version control with context awareness. Each generates/updates context reference files in .ccs/.

Operations

deploy · track · connect

Deployment checklists, session tracking, and MCP server setup.

Agents

Agent File Role Model Tools
context-builder agents/context-builder.md Deep codebase scan → .ccs/ index opus Read, Glob, Grep, Write
test-runner agents/test-runner.md Execute tests, track results, auto-fix sonnet Bash, Read, Grep, Write
code-auditor agents/code-auditor.md Security, perf, dead code, a11y analysis opus Read, Glob, Grep
git-tracker agents/git-tracker.md Branch, PR, merge, diff, sync ops sonnet Bash, Read, Grep, Write
knowledge-guide agents/knowledge-guide.md Methodology guidance, note quality haiku Read, Grep
team-lead agents/team-lead.md Multi-agent coordination, task decomposition opus Read, Write, Glob, Grep, Bash, Task

Dependency Graph

init ──┬──→ status, query, refresh
       ├──→ plan ──→ build
       ├──→ fix, refactor
       ├──→ team (orchestrates plan, build, fix, test, review)
       ├──→ test ──→ fix (auto-fix loop)
       ├──→ audit, review, diff
       ├──→ branch ──→ pr ──→ merge
       ├──→ sync, log, stash
       └──→ deploy

research, track, connect → independent (no init dependency)

Shared State (.ccs/ directory)

File Written By Read By
project-map.md init, refresh all skills
architecture.md init, refresh plan, build, refactor, audit, review
file-index.md init, refresh query, build, fix, refactor
conventions.md init, refresh build, review, audit
task.md all skills (append) status, log, track
preferences.json init refresh
branches/*.md branch pr, merge, sync
pulls/*.md pr merge, review
merge-history.md merge log
commit-log.md log track
team-board.md team team-lead agent

For Swarm Orchestrators

  • Discovery: Read this file (~2.5K tokens) → pick skill → read that skill (~3-8K tokens)
  • Machine-readable: See manifest.json for programmatic access
  • Agent lookup: See AGENTS-INDEX.md for 6 agents with tool restrictions
  • Coordination: Filesystem-based via .ccs/ — no network required
  • Parallel reads: Any number of agents can read the index simultaneously
  • Write lock: Only one agent should write to .ccs/task.md at a time (append-only)
  • Isolation: Each skill operates in forked context — no cross-skill state leakage
  • Model tiering: Use the model column to route to appropriate compute tier

10x-Code — 10x-Code v2.0.0